AIAA 2003-4162

33rd AIAA Fluid Dynamics Conference and Exhibit

23-26 June 2003, Orlando, Florida

EFFECT OF SUB-BOUNDARY LAYER

VORTEX GENERATORS ON INCIDENT TURBULENCE

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

J. Casper*, J. C. Lin, and C. S. Yao

NASA Langley Research Center
Hampton, Virginia

Abstract

Uc
U
u
2
u
u rms
w
v
x
ww

Sub-boundary layer vortex generators were tested

in a wind tunnel to assess their effect on the velocity
field within the wake region of a turbulent boundary
layer. Both mean flow quantities and turbulence
statistics were measured. Although very small relative
to the boundary layer thickness, these so-called micro
vortex generators were found to have a measurable
effect on the power spectra and integral length scales of
the turbulence at a distance many times the height of the
devices themselves. In addition, the potential acoustic
impact of these devices is also discussed. Measured
turbulence spectra are used as input to an acoustic
formulation in a manner that compares predicted sound
pressure levels that result from the incident boundarylayer turbulence, with and without the vortex generators
in the flow.

=
=
=
=
=
=
=
=
=
=
=
=
=
=

local mean flow speed (m/s)

freestream speed (m/s)
streamwise turbulent velocity (m/s)
mean-square streamwise turbulence
u 2 1/ 2
transverse turbulent velocity (m/s)
vertical turbulent velocity (m/s)
[ x1 , x2 , x3 ]T , observer position
2D transverse spectrum (m4/rad2-s2)
c0 / f , acoustic wavelength (m)
ambient density (kg/m3)
source time (s)
random phase variable (radians)
2f , circular frequency (radians/s)

( )

Nomenclature
b
C
c0
f
E uu
E ww
g
h( k c )
kc
P
p
p

=
=
=
=
=
=
=
=
=
=
=
=
r =
t =

1. Introduction

airfoil semi-span (m)

airfoil chord (m)
ambient sound speed (m/s)
frequency (Hz)
1D streamwise spectrum (m3/rad-s2)
1D transverse spectrum (m3/rad-s2)
velocity-to-pressure transfer function
scaling function ( Eqs. (7) and (8) )
/ U c , convective wave number (rad/m)
unsteady surface pressure jump (Pa)
unsteady surface pressure (Pa)
radiated sound pressure radiated (Pa)
distance from source to observer (m)
observer time (s)

Vortex generators (VGs), first introduced by Taylor

[1], have long been used to control boundary-layer
flows by redirecting the outer flow to the wall region
via streamwise vortices [2, 3]. Such devices are also
known to alter the boundary-layer turbulence
downstream of the VGs [4]. The VGs used in those
previous studies generally consisted of a row of small
vanes with a device height approximately the size of the
boundary-layer thickness . The VGs were set at an
angle of incidence to the local flow to produce an array
of streamwise vortices. However, for some flow control
applications, the use of these -scale VGs could
produce excess residual drag through conversion of a
vehicles forward momentum into unrecoverable
turbulence in its wake.

Research Scientist, Computational Modeling and Simulation Branch, AIAA Senior Member.
Senior Research Scientist, Flow Modeling and Control Branch, AIAA Associate Fellow.

Senior Research Scientist, Flow Modeling and Control Branch.

This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.

1
AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS
This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

that turbulence on an airfoil, e.g., a stator beneath the

nacelle model. Such a configuration could be viewed as
a simplified model for the potential control of inlet
distortion downstream of an engine inlet. The flow
measurements, with and without MVGs, are used as
inflow conditions in an acoustic analysis to determine a
sound-pressure-level (SPL) differential. Concluding
remarks are given in Section 5.

One method of reducing the device drag and

improving the system efficiency is to generate an
embedded streamwise vortex using minimal near-wall
protuberances through substantially reduced device
height. Lin et al. [5] showed that, with a reduced device
height in the range of 10 to 20 percent of boundary
layer thickness, these micro vortex generators
(MVGs) could still provide sufficient momentum
transfer to prevent boundary-layer separation. With
proper design guidance, MVGs can efficiently produce
streamwise vortices that prevent separation, yet do not
dominate the boundary layer after the flow-control
objective is achieved [6].

2. Experimental Apparatus
2.1. Facility, Test Set-up, and MVGs

The experimental measurements in this report were

conducted in the NASA Langley Shear Flow Tunnel.
This facility is a subsonic open-circuit wind tunnel with
a 4.67-m long test section that is 50.80 cm by 71.12 cm
(20 x 28 inches) in cross section. A 1.27-cm-thick
splitter plate was mounted 10.16 cm above the original
test section floor to bypass the converging section
boundary layer to eliminate any upstream influence.
The new boundary layer on the splitter plate was
artificially tripped with a 5.08-cm-wide strip of
sandpaper (36 grit) located 25.40 cm from the leading
edge of the splitter plate.

Sub-boundary-layer scale vortex generators have

been successfully demonstrated in a variety of
applications. For instance, such devices were shown to
significantly increase the lift-to-drag ratio on multielement high-lift airfoils [6, 7], or to effectively control
separation in an adverse pressure gradient [8].
However, of even greater interest with respect to the
present study are applications of sub-boundary layer
vortex generators near the entrances and exits of inlet
ducts to enhance the performance of jet engines, e.g.
[9,10,11,12]. Such studies have led to an increased
interest in obtaining information on the potential
acoustic impact of these devices.

A 40.64-cm-chord simulated nacelle model was

installed parallel to the splitter plate with its leading
edge located 3.2 m downstream of the boundary-layer
trip to create a flow-through passage that spanned the
width of the test section (71.12 cm). Schematics of the
experimental set-up in the tunnel and the nacelle model
are illustrated in Figs. 1(a) and 1(b), respectively. The
model is manufactured in three chord-wise pieces: an
aluminum leading edge, a center portion of constant
thickness, and an aluminum trailing edge. The central
portion was made of glass to provide transparency for
particle-image velocimetry (PIV) measurements.
However, it was decided that such measurements would
not be necessary for the current study.

The experimental work reported here was

designed to quantify the effect of MVGs on the
unsteady velocity field within the wake region of a
turbulent boundary layer. A second objective was to
determine the impact that the flow-field alteration has
on the acoustic field downstream of the MVGs.
Thirdly, the authors wanted to determine whether an
economical procedure could be developed for flow
control design studies when acoustics play a roll in the
design space.
The details of the experiment are described in the
following section. A nacelle-like model is installed
near the floor of a low-speed wind tunnel so that the
entire model is immersed in a turbulent boundary layer.
Mean flow and turbulence measurements are taken
between the nacelle model and tunnel floor. MVGs are
attached on the lower side of the nacelle model, near its
leading edge, and the downstream flow field is again
measured. Hot-wire anemometry is used to determine
turbulence spectra which are in turn used to calculate
velocity correlation functions and integral length scales.
The flow field measurements, with and without MVGs,
are reported and compared in Section 3.

The nominal flow speed for all reported test runs is

35.05 m/sec (115 ft/sec). The turbulent boundary layer
just upstream of the nacelle model is fully developed
and has a thickness 4.70 cm. The approximate
Reynolds numbers at this location, relative to distance
downstream of the splitter plate edge and the boundary
layer thickness are, respectively, 7.48 10 6 and
1.10 10 5 . The flow-through passage has an initial
opening gap of 3.23 cm at its leading edge and 3.30 cm
at its trailing edge, as well as a constant gap of 2.67 cm
under the models central section where the flow is
measured.

Section 4 begins with a brief acoustic analysis for

an airfoil in a turbulent stream. The objective of this
section is to determine the downstream effect of the
MVGs on the turbulence as it relates to the incidence of

All data involving MVGs in this report were

generated with delta-wing type MVGs, shown
2

AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS

schematically in Fig. 2. These devices have a vane

height of 2.54 mm (Fig. 2). The MVGs were arranged
in a row of counter-rotating pairs, at alternating incident
angles of 23 degrees, with the pairs spaced 1.27 cm
apart (Fig. 2). The configuration was placed on the
lower surface of the nacelle model within 3.81 cm of the
leading edge. An underside view of the nacelle model
with MVGs installed is shown in Fig. 3.

anemometer bridge output and pitot-static free-stream

velocity can be accurately related through a fourth-order
polynomial curve [13]. The polynomial coefficients
determined from each calibration were used to compute
both the mean and fluctuating velocities from each
respective hot-wire survey. The velocity results were
then entered into a Fast Fourier Transform (FFT)
subroutine that calculated the turbulence spectrum. In
order to maintain the high resolution of the fluctuating
velocities, mean and time-dependent signals were
separated before the A/D conversion and combined for
the total velocity (mean plus fluctuating) computation
and separated again thereafter. As a further assurance
of the hot-wire data accuracy, a temperature sensor on
the ceiling of the tunnel entrance indicated that the
variation in tunnel temperature for any survey was
within 0.28 degree C. The measured velocity error of
the current hot-wire system is estimated to be within
2%, and the hot-wire position is accurate to within
0.04 mm, similar to the hot-wire system reported by
Pack and Seifert [14].

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

2.2. Instrumentation

A single component hot-wire anemometry system

was used to measure the turbulence spectrum in a plane
perpendicular to the mean flow within the flow-through
passage. The hot-wire probe consisted of a platinumplated tungsten wire 0.0005 cm in diameter with an unplated active length of 0.13 cm and operated at an
overheat ratio of 1.8. The probe was connected to a
constant-temperature anemometer (DISA Type 55M01).
The bridge output from the hot-wire anemometer was
acquired by a PC computer through an eight-channel
simultaneous sample, 16-bit analog-to-digital (A/D)
converter (HP E1433A). The sampling rate of 66.67
kHz was used to obtain 100 data blocks, each consisting
of 8192 readings. The signal input was low-pass
filtered at 26 kHz and high-pass filtered at 3Hz.

3. Flow Measurements

All measurements were performed at a nominal

tunnel speed of 35.05 m/sec at each location on a 9-by9-point grid within the flow-through passage as
described in the previous section. Both mean flow
values and turbulence statistics were measured in order
to fully characterize the baseline flow and its alteration
with the use of MVGs. Velocity correlation functions
were calculated from measured turbulence spectra and
in turn were used to compute integral length scales. A
discussion of relevant flow measurements follows.

A two-dimensional traverse system was used to

control the hot-wire probes motion from outside the
test section through two spanwise slots, one on the
tunnel ceiling and the other towards the trailing edge of
the nacelle model. All hot-wire data were taken in a
streamwise perpendicular plane located within the flowthrough passage beneath the transparent section of the
nacelle model (Fig. 1(b)). The coordinate system for
these measurements is shown in Fig. 4. The coordinate
origin is on the tunnel floor directly beneath the midspan location of the nacelle model leading edge and the
x-axis is located along the tunnel floor centerline z = 0 .
A 9 9 grid of survey points was selected within the
plane x = 19.05 cm for the flow field measurements.
Nine vertical points above the splitter plate were chosen
at y = 1.02, 1.52, 1.78, 1.91, 2.03, 2.16, 2.29, 2.41, and
2.51 cm. Nine equally spaced spanwise points were
selected in 1.27 z 1.27 cm. The upper extent of
the data plane was limited to y = 2.51 cm because of the
physical dimension of the probe stem. The lower extent
was truncated at y = 1.02 cm because measurements
below this height showed no significant effect from
MVGs of the size considered here (see Section 3).

3.1. Mean Flow

Mean flow measurements, for the baseline flow and

the flow with MVGs, were inspected at all 81 points
within the data plane for spanwise homogeneity. The
largest deviation from spanwise homogeneity was found
to occur in the flow with MVGs, at the highest vertical
station (y = 2.51 cm) where the flow is most heavily
influenced by the MVGs as well as by the additional
boundary layer on the nacelle model underside. The
magnitude of this velocity deviation was no more than
one m/s, or less than three percent of the tunnel speed.
Therefore, the discussion will focus on the vertical
centerline of the data plane.
Fig. 5 shows measurements of the mean flow speed
U c ( y ) , at the nine vertical stations on the tunnel
centerline ( z = 0 ), for the baseline run and with MVGs
installed, in order to visually compare the two mean
flow states. Clearly, the MVGs have slowed the mean

Immediately prior to and after each hot-wire

survey, the hot-wire probe was moved to a fixed freestream location above the nacelle model to be calibrated
against a nearby pitot-static probe for free-stream
velocities ranging from 3.05 to 36.58 m/sec. The
3

AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS

flow speed as expected, because the MVGs on the

underside of the nacelle models leading edge have
slightly increased the blockage at the passage entrance.
The trend indicated in Fig. 5 is that the MVGs cause
larger reductions in U c ( y ) as the measurement station
approaches the nacelle model underside, with a
maximum reduction of approximately one m/s.

are reduced at wave numbers below approximately 300

m-1 and are increased at higher wave numbers. At y =
2.29 cm and 2.41 cm, the comparative spectra again
become closer as the measurement station enters the
boundary layer on the nacelle model underside, with the
difference being negligible at the highest station y =
2.51 cm, as shown in Fig. 7(d).
The power spectra were post-processed in order to
compute Ruu ( ) , the time correlation function, at each
vertical measuring station. The type of processing used
to calculate Ruu ( ) from Euu (k c ) can be found, e.g., in
[15]. The velocity time-correlations, for the baseline
flow and the flow with MVGs, are plotted in Figs. 8(a)
8(d), for the same measurement stations that
correspond to the PSD plots in Figs. 7(a) 7(d). Not
surprisingly, the velocity time-correlations compare
between the two flow states in a similar manner to the
PSDs. The difference is most significant at the interior
points, but negligible at the uppermost and lowermost
points.

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

3.2. Turbulence Statistics

Similar to the mean flow measurements, the

turbulence measurements were found to deviate most
from spanwise homogeneity at the vertical measurement
station nearest the nacelle model underside, for the flow
with MVGs. The size of the largest deviation in
turbulence intensity was approximately 0.6 percent of
the local mean flow speed.
Fig. 6 shows the local turbulence intensity at z = 0
for the baseline flow and the flow with MVGs. The
local turbulence intensity u c ( y ) is defined by
u c ( y ) = u rms / U c ( y ) ,

u rms = [ u 2 ( y )]1 / 2

If Taylors hypothesis is assumed, these time

correlation functions can be used to determine
streamwise integral length scales in the following way.
First, a temporal scale is calculated by integrating the
correlation function with respect to all time separations
. Then the streamwise integral length scale Luu ( y ) is
obtained by multiplying the temporal scale by the local
mean flow speed:

where u 2 ( y ) is the streamwise local mean-square

turbulence. The MVGs have increased the local
turbulence intensity levels in the upper half of the
measurement plane and decreased the levels in the
lower half. Perhaps the most interesting observation in
Fig. 6 is the vertical distance over which the MVGs
have affected the flow. Although the MVGs protrude
only 2.54 mm from the nacelle model surface, their
effect on the mean turbulence is measurable throughout
the vertical extent of the data measurement plane, i.e.,
approximately 15 mm.

Luu ( y ) U c ( y )

Euu (k c ) dk c = u ( y )
2

Ruu ( ) d

(2)

For present purposes, the integration in Eq. (2) is

performed only for the first positive lobe of the
correlation function. The integral length scales, with
and without MVGs in the flow, are compared in Fig. 9.
Fig. 9 shows that the application of the MVGs has
resulted in a decrease or no change in the streamwise
integral length scale at eight of the nine measurement
locations.

Figs. 7(a) 7(d) show the power spectral density

(PSD) of the streamwise turbulence at four of the nine
vertical measurement stations along z = 0 . The value
plotted is Euu (kc ) , the one-dimensional streamwise
PSD, as a function of the convective wave number
kc = / U c ( y ) . These values Euu (kc ) satisfy the
property

4. Acoustic Analysis

The acoustics of an imaginary airfoil placed in a

stator-like position beneath the nacelle model are
simulated with the use of the flow measurements. An
acoustic source is produced by unsteady lift on the
airfoil. In the acoustic analysis that follows, the
objective is to determine the difference in sound
pressure levels that results when MVGs are installed in
the baseline flow. Therefore, an absolute sound
pressure level is not required for either the baseline flow
or the flow with MVGs, but rather a differential sound
pressure level between the two flow states.

(1)

At y = 1.02 cm, Fig. 7(a), the MVGs are too far

away to have much effect on the power spectrum.
However, at y = 1.78 cm and y = 1.91 cm, the plots in
Figs. 7(b) and 7(c) show a statistically significant
difference between the two flow states. Figs. 7(b) and
7(c) are representative of the five locations at y = 1.52,
1.78, 1.91, 2.03, and 2.16 cm. All five spectra are
similar in that the amplitudes for the flow with MVGs

4
AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS

property is achieved by evaluating the spectral

amplitudes of the surface pressure as a function of the
two-dimensional PSD of the stator-normal velocity
field. If the stator is considered a flat plate, then, as
encountered by the stator surface in the plane z = 0 ,
the pressure jump across the stator surface can be
approximated by

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

4.1. Airfoil in Turbulence

Consider an airfoil that is placed in a stator-like

position beneath the nacelle model with its leading edge
coincident with the data measurement plane as in Fig.
10. The chord length C of the stator is 8.89 cm and its
span 2b is the entire passage height of 2.67 cm. For
convenience, a new coordinate system is established so
that the data plane is located at x = 0, as in Fig. 10. The
airfoil chord now exists on 0 x C and its span on
b y b . A stator in this location is subject to inflow
turbulence as measured in the data plane and discussed
in the preceding section. The turbulent flow field gives
rise to an unsteady pressure distribution on the surface
of the stator and thereby creates a fluctuating lifting
force that is well known to produce noise; see e.g., [16]
and [17]. Both the acoustic source and the noise
produced by this incident turbulence are analyzed in the
following sub-sections.

P( x, y, t ) 2 0U c

An ( y ) e i g ( x; k n ) e i k U t
n

(3)

n= N

k n = n k c ,

n = 0, 1, 2,

, N

k c = k N / N
where 0 is the ambient density and g ( x; k c ) is a
transfer function that is derived from thin airfoil theory
[21]. The upper cutoff wave number k N is chosen
such that the acoustic source amplitude is considered
negligible or is out of measurement range for kc > k N .
The phase angles {n } are independent random
variables uniformly distributed on [0, 2 ] . The factor
of two in Eq. (3) indicates that the pressure is assumed
antisymmetric between the two sides of the stator
surface.

4.2. Broadband Surface Pressure Model

In [18], a surface pressure model for an airfoil in

homogeneous, isotropic turbulence was formulated by
characterizing the velocity field as a linear
superposition of periodic gusts that convect at the
freestream speed over a flat plate. These gusts are
determined by the spectral content of the turbulent
velocity component that is normal to the stator surface.
In the present case, this normal component is in the
direction of the z-axis, i.e., the w component. Although
the boundary-layer turbulence in the present case is not
isotropic and is not homogeneous in the direction of the
y-axis, the stator surface pressure will be modeled
similarly to the analysis in [18] and [19], with a
modification to account for non-homogeneity in the y
direction.

The surface pressure amplitudes An ( y ) are now

discussed. It is argued by Amiet [18] that, within
certain limitations, integration over all spanwise wave
numbers is not required, because only one spanwise
wave number will significantly contribute to the
acoustic signal received by an observer in a fixed
location.
Using a time-domain approach that is
analogous to that of Amiet, it was shown in [19] that
only the zero spanwise wave number was acoustically
relevant for an observer in a spanwise symmetric
location. For an observer in the plane y = 0 , the

The vertical non-homogeneity of the flow will be

accounted for by applying strip theory. The stator
surface is divided into streamwise strips as illustrated in
Fig. 11. Theoretically, the width of a strip is restricted
such that the turbulence within a given strip is
uncorrelated with that in an adjoining strip. Within
each strip, the flow is assumed isotropic with a
streamwise turbulence spectrum that is known from
measurement, and the analysis in [19] is applied locally,
with a final integration along the y-axis.

amplitudes An ( y ) are determined by

1

An ( y ) = [ ww ( y; k n ,0) k c ] 2

(4)

where
the
local
two-dimensional
spectrum
ww ( y; kc , k y ) is interpreted as a three-dimensional
PSD of w , into which all stator-surface-normal wave
numbers are integrated:

ww ( y; k c , k y ) =

The evaluation of the stator-surface pressure jump

is accomplished by first recognizing the incident
turbulent fluctuations as a stochastic process. Such a
process can be characterized by a truncated series
whose coefficients are chosen such that the
autocorrelation of the series forms a Fourier transform
pair with its power spectrum (see, e.g., [20]). This




ww ( y; k c , k y , k z ) dk z

(5)

It remains to determine ww ( y; kc ,0) , the zero-th

spanwise wave-number component of the twodimensional transverse velocity spectrum in Eq. (5).
Recall that the streamwise spectrum was measured in
the present experiment and, therefore, the transverse
spectrum must be determined from it.
5

AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS

layer, would collapse when scaled on inner variables.

Fig. 13 shows the collapsed results for the spectral
densities of all three velocity components, as interpreted
in [25]; see Fig. 5-38 in that reference. The values
plotted in Fig. 13 are the one-dimensional streamwise
and transverse power spectra, scaled by 0U c / w y ,
where w is the wall shear stress. The spectra are given
as functions of the dimensionless quantity y / U c .
Recall that U c is a function of y.

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

4.3. Transverse Velocity Spectrum

The transverse velocity spectrum ww (k c ,0) that

is required in Eqs. (3) and (4) is not immediately
available, and must be determined from the available
measurements. There are two steps in this evaluation.
First, the one-dimensional transverse velocity PSD,
E ww (k c ) , is related to the measured streamwise PSD.
Then, an expression is derived that relates the onedimensional transverse spectrum to ww (k c ,0) . Note
that, for the present discussion, the y dependence in the
flow is suppressed, i.e., the stator surface pressure is
modeled within a single strip (Fig. 11).

An ad hoc function h(kc ) is now desired that will

provide a scaling relationship between the streamwise
and transverse spectra, i.e.

If the turbulence were isotropic, then the onedimensional transverse spectrum could be directly
related to the measured, one-dimensional streamwise
spectrum by (see e.g., [22])

E ww (k c ) h(k c ) E uu (k c )

Fig. 14 shows an example of the application of the

inner-variable scaling in Fig. 13 for the present case,
using specific values of y = 1.78 cm and U c = 31.91
m/sec, which correspond to one of the present
measurement stations. The scaled spectra in Fig. 14 are
plotted as a function of kc . The darkened symbols in
Fig. 14 represent the values of the streamwise spectrum,
having been scaled by a function h(kc ) of the form

1
d
Euu (kc )
Eww (kc ) =
Euu (kc ) kc
2
dkc


(7)

(6)

It has been previously demonstrated [23] that Eq.

(6) will result in significant error for a transverse
spectrum in a boundary layer. Therefore, a single
acoustic prediction that incorporates Eq. (6) would also
contain significant error, particularly in the lower
frequencies, where incident turbulence noise dominates.
However, because the desired acoustic prediction is a
differential, the result will ultimately rely on a
difference of PSDs between the two flow states.
Therefore, it could be rationalized that the error made
by Eq. (6), as applied to both flow states, would cancel
in the final calculation of this differential. There is
another problem, however, in the use of Eq. (6) directly
on the measured PSD, as represented in Figs. 7(a)
7(d). The application of the derivative operator to nonsmooth data gives wildly fluctuating results as kc
becomes large, as shown in Fig. 12. Consideration was
given to smoothing either the measured data or the
computed transverse spectrum, but this idea was
abandoned on the basis that the final acoustic prediction
should not be dependent on the characteristics of a
given smoothing algorithm, especially in this case,
where the amount of smoothing must increase as the
wave number increases. Therefore, a different approach
is taken that will allow the use of the measured data as
is.

agreement between the product h(k c ) Euu and the innervariable-scaled transverse spectrum, i.e. Eq. (7) is
satisfied. It was found that, when using all current
measurement stations on 1.02 cm y 2.51 cm , with
corresponding convection speeds U c ( y ) from Fig. 5,
similar plots to Fig. 14 were generated and found to
satisfy the relationship in Eqs. (7) and (8). The values
of the parameter y were required to range from 0.255

In the work of Bradshaw and Ferris [24],

measurements were taken for the PSD of all three
velocity components, as a function of the streamwise
wave number, in a turbulent boundary layer. It was
shown that the measured spectra for a given velocity
component, at different locations within the boundary

Although this approach is ad hoc, again, the

rationale is that the error made in this approximation
will cancel in the final differential calculation, provided
that h(k c ) is a relatively weak function in y, as is

h(k c ) = 1 y log 10 (k max / k c ) ,

0 < k c < k max

(8)

where k max is chosen so that f max = 10 kHz and

y = 0.270 for the case in Fig. 14. Note the close

to 0.290 in order to establish a specific scaling function

h(kc ) for each measurement station. The values of y
for all nine stations are shown in Table I. Note that y
is nearly constant with values ranging from 0.268 to
0.275 on 1.52 cm y 2.16 cm .
The measured
difference between the baseline spectra and the spectra
with MVGs was most significant within this vertical
span of the data measurement plane (See Figs. 7(b) and
7(c).).

6
AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

suggested by the relatively uniform values of y

Now, the autocorrelation R ww (k c ,0) in Eq. (11) is

the equivalent of E ww (k c ) because it represents a onedimensional PSD in wave number space. This is readily
apparent by setting = 0 in Eq. (10):

on the inner measurement stations where the spectra

were most affected by the MVGs. Furthermore, the
measured data for E uu (k c ) can be used as is, without
concern for pre- or post-processing. As an example
application to current data, Fig. 15 shows the computed
values of E ww (k c ) , using the measured streamwise
spectrum E uu (k c ) at y = 1.78 cm and the scaling law in
Eqs. (7) and (8). In order to provide a sanity check for
the scaling method in Eqs. (7) and (8), the computed
values of the baseline transverse spectra are integrated
to calculate a transverse mean-square turbulence, i.e.,




E ww (k c ) dk c = w 2 ( y )

Rww (k c ,0) =

(9)




ww (k c , k y ) e

i k y

dk y

( ) =
=

1
R ww (k c ,0)

Rww (k c ,0)

ww (k c , k y ) dk y E ww (k c )

(12)

1


( ) h(k c ) E uu (k c )

(13)

where h(k c ) is given by Eq. (8).

If the flow is
assumed isotropic within a strip, then the correlation
length y ( ) can be calculated by


( ) =

8 Luu (1 / 3)
3
(5 / 6 )

(14)
2




(k c / k E ) 2

[3 + 8 (k c / k E ) 2 ][1 + (k c / k E ) 2 ]1 / 2


where ( ) denotes the well-known Gamma function

(see, e.g., [26]), and k E is the peak wave number and is
approximately k E 0.747 / Luu . See [18] for a
derivation of Eq. (14).
Eq. (13) is an important result. Whenever possible,
it is desirable to use a streamwise, one-dimensional
spectrum because the quantity E uu (k c ) is typically the
simplest spectrum to measure in a turbulent flow, and
can be accomplished with a single hot-wire element.
Eqs. (13) and (14) now enable the evaluation of the
stator surface pressure in Eq. (3). Such approximations
as discussed above for the modeling of surface pressure
in non-homogeneous, anisotropic turbulence are
admittedly crude. However, the formulation for the
desired differential in sound pressure levels will
ultimately rely on the ratio of two acoustic powers in
which significant cancellation occurs, as discussed in
Section 4.5.

(10)

If R ww (k c , ) is integrated in , then the result is

proportional to a spanwise correlation length. In fact, a
stator-spanwise correlation length y ( ) can be defined

ww (k c ,0) =

The one-dimensional transverse PSD, E ww (k c ) ,

must now be related to ww (k c ,0) , the PSD that is
required to model the stator surface pressure in Eqs. (3)
and (4). This can be accomplished as argued by Amiet
in [18]. Amiets argument relies on the assumption of
isotropy, which is interpreted in the present context as
isotropic within a strip. First, consider the quantity
R ww (k c , ) , a cross-PSD of the transverse velocity
fluctuations, whose values are correlated through a
stator-spanwise distance , i.e.





Eqs. (11) and (12) relate the one-dimensional

transverse PSD to the spectrum ww (k c ,0) that is
required in the stator surface pressure model. Finally,
using Eq. (7), ww (k c ,0) is related to the measured
streamwise PSD by

The values in Eq. (9) are used to determine a

transverse turbulence intensity wrms / U at each
vertical measurement station, as shown in Table I. The
ratios of streamwise and transverse turbulence
intensities are reasonably consistent with the turbulence
intensity data of Klebanov [23] in the boundary layer
region 0.22 y / 0.53 (See Figure 4 in [23].). The
computed vales in Table I are not expected to be
identical to the data in [23], as the present experiment
does not deal with a simple flat plate boundary layer,
but one that contains an obstruction.

Rww (k c , ) =

4.4 Acoustic Formulation

Rww (k c , ) d

The desired noise level comparison will be

accomplished with an acoustic formulation introduced
in [27].
This time-domain formulation, called
Formulation 1A, is a solution of the Ffowcs WilliamsHawkings equation [28]. For the present case, the
observer location x will be assumed to be at a distance
r directly outward from the geometric center of the

(11)
ww (k c ,0)

The second equality in Eq. (11) follows from the

Fourier transform relation between Rww and ww [18].

7
AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS

stator surface, i.e., in the plane y = 0 . Furthermore, the

Mach number of the mean flow is small ( M 0.1) and
the observer location is presumed in the acoustic far
field, as well as the geometric far field, i.e, the
following conditions are satisfied:
r >> ,

M << 1 ,

r >> C

locations { y j }9j =1 .

To this end, the far field noise

associated with the unsteady surface pressure on the j-th

strip is denoted p j ( x , t ) , and is defined by

p ( x , t )

 

(15)

where = c 0 / f is the largest acoustic wavelength of

interest, and c 0 is the sound speed. Under such
conditions, Formulation 1A allows the acoustic pressure
p ( x , t ) received by the observer to be approximated by
a simple surface pressure integration, which can be
written

j =1 n = N

2 c0 r

i U c2 k n I n An ( y j ) e i ( k

U c + n )

p j ( x , t )

(21)

j =1

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

4 p( x , t )

c0 r

C
0

p ( ) dx dy

where j is the discrete stator-spanwise spacing

determined by the flow measurement locations.
The quantity to be compared is the sound pressure
level (SPL). The SPL associated with the acoustic
source on the j-th strip is defined by

(16)

where r is the mean observer distance to the stator

surface, and p ( ) is the pressure on the stator surface
evaluated at a mean retarded time, i.e.,

p( ) = P( x, y, ) ,

= t r / c0

SPL j ( f n ) = 10 log

b
b

0U

2
c

2 c0 r

An ( y ) i k n g ( x; k n ) e i ( k

j-th strip at the frequency f n . The over-bar denotes a

time average, and Pref is a reference pressure most
commonly taken to be 20 Pa. The values of the
pressure spectrum {Pn , j } are the resulting amplitudes of

U c + n )

a Fourier analysis of the time series in Eq. (21) at a

point y j on b < y j < b . The total SPL is obtained by

dx dy

n= N

power-summing the contributions due to the component

sources on each strip:

For convenience, the acoustic signal in Eq. (18) is

rewritten as

p ( x , t )

2 c0 r

b n= N

iU c2 k n I n An ( y ) e i ( k

U c + n )

SPL ( f n ) = 10 log

dy

where

p j (t ) =

g ( x; k n ) dx

   

10

0.1 SPL j ( f n )

j =1




(23)

The far field acoustic pressure due to the source on

the j-th strip in Eq. (21) can be written in the form

(19)

In =

(22)

Pref2

where Pn2, j is the acoustic PSD due to the source on the

(18)
N

Pn2, j

(17)

If the surface pressure jump in Eq. (3) is substituted in

Eqs. (16) and (17), the result is the following expression
for the acoustic pressure signal at the observer in the far
field:

p ( x , t )

  
 

i
q n, j e

n, jt

(24)

n= N

(20)

where n = k nU c . It can be shown (see, e.g., [29]) that

A formula for the integral in Eq. (20) is given in [21];

see Equations 18 and 19 of that reference.

the time average of [ p j (t )] 2 can be evaluated as

[ p j (t )] 2 = [ p j (t )][ p j (t )]* =

4.5 Incident Turbulence Noise Comparison

n= N

Baseline noise levels and noise levels that include

the effect of MVGs are now compared. These noise
levels are based on acoustic pressure signals as
evaluated by Eq. (19). The numerical evaluation of the
spanwise integral in Eq. (19) requires a discretization of
the stator span according to the nine flow measurement

q n, j

Pn2, j

(25)

n =0

where the star superscript a denotes complex conjugate.

Eqs. (22), (23), and (25) allow the total SPL at a given
non-negative frequency to be written

8
AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS

Recall that k E 1 / Luu . Now, if Eqs. (4), (13), and

SPL( f n ) = 10 log

2
Pref

q n, j

j =1

(26)

(14) are substituted for An2 in Eq. (28) and I n , h(k c ) ,

and
and

q n , j

= q n , j q

( ) are nearly constant in y for a fixed

frequency, then some algebraic manipulation yields the

following simplification of Eq. (28):

*
n, j

9


02
2
o

4c r

[U

4
c

2
n

2
n j

In k A

(27)


2
j

SPL( f n ) = 10 log




j =1
9

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

j =1

Note that the surface pressure amplitudes { An } are real.

In order to compare noise levels arising from the two
flow states, let the subscript notation ( ) B and ( ) V
represent quantities that are evaluated with respect to
the baseline flow and the flow with MVGs, respectively.
Using Eqs. (26) and (27), the desired SPL differential
induced by the entire stator source is evaluated by

= 10 log













j =1











j =1

{ [U
{ [U

In k A

4
c

I n k n2

2
n

]}
A ] }

4
c

2
n j

2
n j

2
j

2
j

(28)








where An is given in Eqs. (4), (13), and (14).

Fig. 16 is a plot of Eq. (28), using the current
measurements at the nine stations along z = 0, as a
function of frequency. The application of the MVGs
has resulted in an SPL differential in the overall range
of approximately 2 dB out to 27 kHz. Note the
average relative decrease in SPL below approximately
two kHz.
For present purposes, Eq. (28) can be further
simplified by the following reasoning. First, the value
of I n is relatively insensitive to the convection speed
for low Mach number flows [21], and therefore is nearly
constant throughout the data measurement plane, for a
given frequency. The ad hoc scaling function h(k c )
that occurs Eq. (13) is also nearly constant in y for a
fixed frequency, in particular for those locations at
which the MVGs have their largest impact on the
turbulence spectrum. Furthermore, the stator-spanwise
correlation length y ( ) is virtually independent of y

( )

Luu (k c / k E ) 2
1

2
kc
(kc / k E ) (k c / k E )

Euu (k n )]

(30)


Micro vortex generators (MVGs) have been tested

in a wind tunnel to quantitatively assess their effect on
the velocity field within the wake region of a turbulent
boundary layer. Measurements of the mean flow and
turbulence indicated that the MVGs had a measurable
effect on the flow, relative to the baseline. Velocity
power spectra and integral length scales of the
turbulence were altered at a vertical distance several
times the height of the MVGs. The application of the
MVGs resulted in a reduction in power spectral
amplitude at frequencies below approximately 1500 Hz,
and an increase at higher frequencies. In addition, the
streamwise integral length scales were reduced by as
much as 14 percent. A decrease in streamwise length
scales suggests that the MVGs have reorganized the
turbulence in a more isotropic fashion, although

for all but the lowest wave numbers as can be seen by

the following proportionality when k c >> 1 in Eq. (14):
y

2
j Bj



5. Concluding Remarks

Euu (k n )]

As a final calculation, upper and lower bounds of

the SPL differential are established by using data from a
measurement station in which the difference in
turbulence spectra is most prominent, say at y = 1.91
cm. Fig. 18 shows an SPL differential as predicted with
the assumption that the mean flow and turbulence
spectrum as measured at y = 1.91 cm is held constant
throughout the vertical extent of the measurement plane.
Under this worst case scenario, the SPL differential
range has been extended from approximately 3 dB to 5
dB. The reasoning behind such a calculation is to help
determine if an economical procedure can be developed
for flow control design studies, i.e., when only one
measurement is affordable for each point in a design
matrix.

{ [U

Note the significant simplification in Eq. (30) that

occurs as a result of cancellation, owing to the
differential formulation. The SPL differential, as
predicted by Eq. (30) with the measured input, is shown
in Fig. 17.
The plot in Fig. 17 is visually
indistinguishable from Fig. 16, thereby requiring
separate figures for their comparison.

SPL( f n ) [SPL( f n )]V [SPL( f n )]B



{ [U

2
j V j

(29)

9
AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS

9.

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

additional measurements would be necessary to confirm

this.
The potential acoustic impact of the MVGs was
analyzed. The acoustic analysis resulted in a relatively
simple formulation that was used to determine the
additional noise that such devices impart to a turbulent
flow. Measured turbulence spectra were input to the
acoustic formulation to predict the effect of the MVGs
on radiated noise that results from incident turbulence.
The acoustic analysis indicated that the MVGs would
cause a relatively small change in sound pressure levels
of approximately 2 dB. Acoustic measurements would
be required to fully validate this analytical approach.

10. Anderson, B. H. and Gibb, J., Vortex-Generator

Installation Studies on Steady-State and Dynamic
Distortion, AIAA Journal of Aircraft, Vol. 35, No.
4, 1998, pp. 513 520.
11. Anderson, B. H., Miller, D. N., Yagle, P. J., and
Truax, P. P., A Study on MEMS Flow Control for
the Management of Engine Face Distortion in
Compact Inlet Systems, ASME Paper FEDSM999620, Proceedings of the 3rd ASME/JSME Joint
Fluids Engineering Conference, July 1999.

References

1.

Taylor, H. D., The Elimination of Diffuser

Separation by Vortex Generators, United Aircraft
Corporation Report No. R-4012-3, June 1947.

2.

Schubauer, G. B., and Spangenber, W. G., Forced

Mixing in Boundary Layers, Journal of Fluid
Mechanics, Vol. 8, Part 1, 1960, pp. 10 32.

3.

Pearcey, H. H., Shock Induced Separation and Its

Prevention by Design and Boundary-Layer
Control, Boundary Layer and Flow Control, Vol.
2, ed. G. V. Lachman, Pergamon Press, New York,
1961, pp. 1166 1344.

4.

Shaw, R. J., An Experimental Investigation of

Force Mixing of A Turbulent Boundary Layer in an
Annular Diffuser, NASA TM 79171, April 1979.

5.

Lin, J. C., Howard, F. G., and Selby, G. V., Small

Submerged Vortex Generators for Turbulent Flow
Separation Control, Journal of Spacecraft and
Rockets, Vol. 27, No. 5, Sept.-Oct. 1990, pp. 503
507.

6.

12. Sullerey, R. K., Mishra, S., and Pradeep, A. M.,

Application of Boundary Layer Fences and Vortex
Generators in Improving Performance of S-Duct
Diffusers, Journal of Fluids Engineering, Vol.
124, 2002, pp. 136 142.
13. Sandborn, V. A., Resistance Temperature
Transducers, Metrology Press, Fort Collins, 1972.
14. Pack, L. G., and Seifert, A., Multiple Mode
Actuation of a Turbulent Jet, AIAA Paper 010735, 39th AIAA Aerospace Sciences Meeting and
Exhibit, Reno, NV, January 8-11, 2001.
15. Bendat, J. S. and Piersol, A. G., Random Data:
Analysis and Measurement Procedure, second
edition, John Wiley and Sons, New York, 1986.
16. Lamb, H., Hydrodynamics, New York, Dover
Publications, 6th edition, 1932, p. 501.
17. Curle, N., The Influence of Solid Boundaries upon
Aerodynamics Sound, Proceedings of the Royal
Society A231, 1954, pp. 505 514.

Lin, J. C., Control of Turbulent Boundary-Layer

Separation Using Micro-Vortex Generators, AIAA
Paper 99-3404, 30th AIAA Fluid Dynamics
Conference, Norfolk, VA, June 28-July 1, 1999.

7.

Lin, J. C., Robinson, S. K., McGhee, R. J., and

Valarezo, W. O., Separation Control on High-Lift
Airfoils Via Micro-Vortex Generators, Journal of
Aircraft, Vol. 31, No. 6, Nov.-Dec. 1994, pp. 1317
1323.

8.

Jenkins, L., Gorton, S. A., and Anders, S., Flow

Control Device Evaluation for An Internal Flow
with An Adverse Pressure Gradient, AIAA Paper
2002-0266, 40th AIAA Aerospace Sciences
Meeting and Exhibit, Reno, NV, January 14-17,
2002.

Hamstra, J. W., Miller, D. N., Truax, P. P.,

Anderson, B. H., and Wendt, B. J., Active Inlet
Flow Control Technology Demonstration, ICAS2000-6.11.2, 22nd International Congress of the
Aeronautical Sciences, Aug.-Sept. 2000.

18. Amiet, R. K., Acoustic Radiation from an Airfoil

in a Turbulent Stream, Journal of Sound and
Vibration, Vol. 41, 1975, pp. 407 420.
19. Casper, J. and Farassat, F., A New Time Domain
Formulation for Broadband Noise Predictions,
International Journal of Aeroacoustics, Vol. 1, No.
3, 2002, pp. 207 240.
20. Shinozuka, M. and Deodatis, G., Simulation of
Stochastic Processes by Spectral Representation,
Applied Mechanics Review, Vol. 44, No. 4, 1991,
pp. 191 204.
21. Amiet, R. K., High Frequency Thin-Airfoil Theory
for Subsonic Flow, AIAA Journal, Vol. 14, No. 8,
1976, pp. 1076-1082.

10
AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS

22. Hinze, J. O., Turbulence, second edition, McGrawHill, 1975.

U

23. Klebanov, P. S., Characteristics of Turbulence in a

Boundary Layer with Zero Pressure Gradient,
NACA Report 1247, 1955.

Nacelle
Trip

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

24. Bradshaw, P. and Ferriss, D. H., The Response of

a Retarded Equilibrium Turbulent Boundary Layer
to the Sudden Removal of Pressure Gradient,
National Physics Lab Aero Report 1145, 1965.

Splitter Plate

Tunnel Floor

(a) Test set-up in wind tunnel.

25. White, F. M., Viscous Fluid Flow, McGraw-Hill,

1974.

3.81 cm
24.13 cm

26. Abramowitz, M. and Stegun, I. A., Handbook of

Mathematical Functions, Dover, 1972.

Data
plane

MVG

27. Farassat, F. and Succi, G. P., The Prediction of

Helicopter Rotor Discrete Frequency Noise,
Vertica, Vol. 7, No. 4, 1983, pp. 309 320.

12.70 cm

2.67 cm

19.05 cm

(b) Nacelle model dimension.

28. Ffowcs Williams, J. E. and Hawkings, D. L.,

Sound Generation by Turbulence and Surfaces in
Arbitrary Motion, Philosophical Transactions of
the Royal Society, A 264, 1969, pp. 321 342.

Figure 1. Schematics for experimental set-up.

29. Pierce, A. D., Acoustics, Acoustical Society of

America, New York, 1991.

Top View

Side View

h
23o

y
(cm)

Uc
(m/s)

1.02

29.76

0.290

5.86

4.23

0.72

1.52

31.17

0.275

5.32

3.93

0.74

1.78

31.91

0.270

5.08

3.79

0.75

1.91

32.25

0.270

4.86

3.66

0.75

2.03

32.50

0.270

4.72

3.54

0.75

2.16

32.69

0.268

4.67

3.52

0.76

2.29

32.29

0.265

4.84

3.73

0.77

2.41

30.21

0.260

5.88

4.70

0.80

2.51

26.66

0.255

6.51

5.37

0.83

u rms /U wrms /U wrms /urms

(%)
(%)

Figure 2. Delta-wing MVG geometry.

MVGs

Table I. Scaling values and turbulence intensities for

baseline flow state.
Figure 3. Underside view of nacelle model with MVGs.

11
AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS

10-1

-2

10

x
10-3

Euu

1.02cm
cm
yyyy==1.02

-4

10

Baseline
Baseline
MVGs
MVGs

0
-5

10

z
1

10

10

10

kc

Figure 4. Experimental coordinate system.

Figure 7(a). Turbulence power spectra at y = 1.02 cm.

10-1

10-2

10-3

Baseline

Euu

y, cm , cm
Vertical Location

2.5

MVGs

1.5

1.78cm
cm
yyyy==1.78

10-4

Baseline
Baseline
MVGs
MVGs
-5

10

1
26

28

30

32
10-6

UcU
(y) , m/s

101

102

103

kc

Figure 5. Mean flow at z = 0.

Figure 7(b). Turbulence power spectra at y = 1.78 cm.

10-1

2.5

10-2

2
10-3

Euu

y, cm , cm
Vertical Location

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

-6

10

Baseline
MVGs

1.5

1.91cm
cm
yyy==1.91

10-4

Baseline
Baseline
MVGs
MVGs
-5

10

1
4

10-6

10

101

102

103

kc

u c(y)
/U , percent
urms/U

Figure 7(c). Turbulence power spectra at y = 1.91 cm.

Figure 6. Local turbulence intensity at z = 0.

12
AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS

1.0
10-1

y = 1.91 cm
0.8
Baseline
MVGs

10-2

0.6

Euu

Ruu

10-3

2.51cm
cm
yyy==2.51

10-4

0.4

Baseline
Baseline
MVGs
MVGs

0.2

-5

0.0
10-6

101

102

103

-5

-4

10

kc

Figure 7(d). Turbulence power spectra at y = 2.51 cm

10

-3

-2

10

10

Figure 8(c). Velocity time correlations at y = 1.91 cm.

1.0

1.0

y = 1.02 cm

y = 2.51 cm

0.8

0.8
Baseline
MVGs

Baseline
MVGs

Ruu

0.6

Ruu

0.6

0.4

0.4

0.2

0.2

0.0

0.0
-5

10

-4

10

-3

10

-2

-5

10

-4

10

Figure 8(a). Velocity time correlations at y = 1.02 cm.

10

-3

-2

10

10

Figure 8(d). Velocity time correlations at y = 2.51 cm.

1.0

y = 1.78 cm
0.8
Baseline
MVGs

cm

2.5

Luu ,

0.6

Ruu

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

10

0.4

1.5

Baseline

0.2

MVGs

1
0.0
-5

10

-4

10

-3

10

0.5

-2

10

1.5

y,

Figure 8(b). Velocity time correlations at y = 1.78 cm.

2.5

cm

Figure 9. Streamwise integral length scales.

13
AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS

1 9 .0 5 c m

M VG

D a ta
p la n e

2 .6 7 c m

y = b

s ta to r

x
y = -b
C

Figure 10. Placement of imaginary stator with leading edge at data measurement plane. Coordinate origin moved to stator
leading edge for acoustic analysis.
103

10

10

u 2 ( ) 0 U c / w y
v 2 ( ) 0 U c / w y

y
x

w 2 ( ) 0 U c / w y

100

Figure 11. Strip theory schematic for sstator surface

pressure. Isotropy within a strip is assumed in the analysis.

10

-1

10

-2

10

-3

10-2

10-1

100

y /Uc

101

102

Figure 13. Streamwise, wall-normal, and transverse

velocity spectra, scaled on inner variables. Data of
Bradshaw and Ferriss [24], as interpreted and reproduced
from White [25].
101

-1

10

100

10-2
10-1

Euu , Eww

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

( T u n n e l F lo o r )

10-3

10-2

u 2 0 U c / w

10-3

10-4

v 2 0 U c / w
w 2 0 U c / w

10-4

Euu , measured

-5

10

h( k c ) u 2 0 U c / w

Eww , isotropy relation

-5

10

100

101

-6

10

10

10

10

kc

102

103

Figure 14. Scaled spectra from Fig. 13 with y = 1.78 cm

and U c = 31.91 m/s. Dark symbols are the streamwise

kc

Figure 12. Streamwise spectrum as measured at y = 1.02

cm, and transverse spectrum as computed by the isotropic
relation in Eq. (6).

spectrum scaled by h(kc) in Eq. (8) with y = 0.270 .

14

AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS

-1

10

10-2

SPL , dB

10-3

-4

10

-2

Euu , measured

-5

10

Eww , computed

-4

-6

10

10

10

10

10

10

15

20

25

Frequency , kHz

kc

Figure 17. Difference in SPLs between the baseline flow

and the flow with MVGs, as predicted by Eq. (30) with
experimental input.

Figure 15. Measured streamwise PSD , at y = 1.78 cm;

transverse PSD computed with scaling laws in Eqs. (7) and
(8), and y = 0.270 .

SPL , dB

SPL , dB

Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 4, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2003-4162

Euu , Eww

-2

-2

-4

-4
0

10

15

20

25

10

15

20

25

Frequency , kHz

Frequency , kHz

Figure 16. Difference in SPLs between the baseline flow

and the flow with MVGs, as predicted by Eq. (28) with
experimental input.

Figure 18. Worst case SPL differential, as predicted by

Eq. (30) with experimental input only from measurement
station y = 1.91 cm.

15
AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS